Welcome to Building Coding Agents with Tool Execution, both in partnership with E2B and taught by Tereza Tizkova and Francesco Zuppichinni. In this course, you learn how to build agents that can generate and execute code as part of how they get the work done. Many of us will predefine some set of functions that we give to, say, an LLM to call. But rather than using only a predefined small set of functions, if you instead let your LLM or your agent write and execute arbitrary Python code. then has access to every built-in Python function or can even ask to install common Python packages and call functions from that. This really opens up the space of what an agent can now do. Also, instead of sequentially making function calls one at a time, which is common in LLM tool use for slightly more complex tasks, your agent can now also just write a huge chunk of code. and efficiently execute a long sequence of operations, and this too makes it more capable. But running code generated by an LLM on your local machine can be unsafe. What if the code does something it isn't supposed to? Once, a highly agentic coding tool deleted one of my team members' important project files. And another did a flawed database migration that deleted production data. Thankfully, we had backups. But to run generated code safely, a good practice is to run that code inside a sandbox environment so that the code execution is isolated from your system. Thanks Andrew. In this course, we will start by breaking down what an AI coding agent really is and how it differs from other AI agents. You will learn how an LLM can reason step-by-step, decide which tool to call, execute code, and use the result to plan its next action. And then, you'll build your own agent capable of running Python code, reading and writing files and iterating through tasks until it reaches a final answer. We'll discuss how to manage the context window, handle errors, and retries. Once you understand how an agent thinks and acts, you'll explore different environments like local execution, containers, and secure sandboxed microVMs. And you will learn how to choose the right one. depending on your use case. You'll learn how to create and manage sandboxes, run code remotely in different languages and interact with the sandbox file system. So your agent will be able to use uploaded dataset, save outputs and even create and host small web application. You build a data analyst agent that explores CSV files with pandas and generates clear visualization. And in a final lesson, we'll create a full-stack agent that edits multiple files and builds a working Next.JS web app inside the sandbox. To keep your agent focused even across large, multifile projects, we'd also discuss managing long context through runtime summarization. Many people have worked to create this course. I'd like to thank Jakub Dobry from E2B and from DeepLearning.AI, Esmaeil Gargari also contributed to this course. The first lesson will introduce you to the inner workings of coding agents, as well as some challenges in building them. So, let's go on to the next video and get started.